2 * This file is part of the GROMACS molecular simulation package.
4 * Copyright (c) 2012,2013,2014, by the GROMACS development team, led by
5 * Mark Abraham, David van der Spoel, Berk Hess, and Erik Lindahl,
6 * and including many others, as listed in the AUTHORS file in the
7 * top-level source directory and at http://www.gromacs.org.
9 * GROMACS is free software; you can redistribute it and/or
10 * modify it under the terms of the GNU Lesser General Public License
11 * as published by the Free Software Foundation; either version 2.1
12 * of the License, or (at your option) any later version.
14 * GROMACS is distributed in the hope that it will be useful,
15 * but WITHOUT ANY WARRANTY; without even the implied warranty of
16 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
17 * Lesser General Public License for more details.
19 * You should have received a copy of the GNU Lesser General Public
20 * License along with GROMACS; if not, see
21 * http://www.gnu.org/licenses, or write to the Free Software Foundation,
22 * Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA.
24 * If you want to redistribute modifications to GROMACS, please
25 * consider that scientific software is very special. Version
26 * control is crucial - bugs must be traceable. We will be happy to
27 * consider code for inclusion in the official distribution, but
28 * derived work must not be called official GROMACS. Details are found
29 * in the README & COPYING files - if they are missing, get the
30 * official version at http://www.gromacs.org.
32 * To help us fund GROMACS development, we humbly ask that you cite
33 * the research papers on the package. Check out http://www.gromacs.org.
36 * Note: this file was generated by the GROMACS sse2_double kernel generator.
42 #include "../nb_kernel.h"
43 #include "types/simple.h"
44 #include "gromacs/math/vec.h"
47 #include "gromacs/simd/math_x86_sse2_double.h"
48 #include "kernelutil_x86_sse2_double.h"
51 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomW3P1_VF_sse2_double
52 * Electrostatics interaction: ReactionField
53 * VdW interaction: LennardJones
54 * Geometry: Water3-Particle
55 * Calculate force/pot: PotentialAndForce
58 nb_kernel_ElecRF_VdwLJ_GeomW3P1_VF_sse2_double
59 (t_nblist * gmx_restrict nlist,
60 rvec * gmx_restrict xx,
61 rvec * gmx_restrict ff,
62 t_forcerec * gmx_restrict fr,
63 t_mdatoms * gmx_restrict mdatoms,
64 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
65 t_nrnb * gmx_restrict nrnb)
67 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
68 * just 0 for non-waters.
69 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
70 * jnr indices corresponding to data put in the four positions in the SIMD register.
72 int i_shift_offset,i_coord_offset,outeriter,inneriter;
73 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
75 int j_coord_offsetA,j_coord_offsetB;
76 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
78 real *shiftvec,*fshift,*x,*f;
79 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
81 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
83 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
85 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
86 int vdwjidx0A,vdwjidx0B;
87 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
88 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
89 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
90 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
91 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
94 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
97 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
98 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
99 __m128d dummy_mask,cutoff_mask;
100 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
101 __m128d one = _mm_set1_pd(1.0);
102 __m128d two = _mm_set1_pd(2.0);
108 jindex = nlist->jindex;
110 shiftidx = nlist->shift;
112 shiftvec = fr->shift_vec[0];
113 fshift = fr->fshift[0];
114 facel = _mm_set1_pd(fr->epsfac);
115 charge = mdatoms->chargeA;
116 krf = _mm_set1_pd(fr->ic->k_rf);
117 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
118 crf = _mm_set1_pd(fr->ic->c_rf);
119 nvdwtype = fr->ntype;
121 vdwtype = mdatoms->typeA;
123 /* Setup water-specific parameters */
124 inr = nlist->iinr[0];
125 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
126 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
127 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
128 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
130 /* Avoid stupid compiler warnings */
138 /* Start outer loop over neighborlists */
139 for(iidx=0; iidx<nri; iidx++)
141 /* Load shift vector for this list */
142 i_shift_offset = DIM*shiftidx[iidx];
144 /* Load limits for loop over neighbors */
145 j_index_start = jindex[iidx];
146 j_index_end = jindex[iidx+1];
148 /* Get outer coordinate index */
150 i_coord_offset = DIM*inr;
152 /* Load i particle coords and add shift vector */
153 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
154 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
156 fix0 = _mm_setzero_pd();
157 fiy0 = _mm_setzero_pd();
158 fiz0 = _mm_setzero_pd();
159 fix1 = _mm_setzero_pd();
160 fiy1 = _mm_setzero_pd();
161 fiz1 = _mm_setzero_pd();
162 fix2 = _mm_setzero_pd();
163 fiy2 = _mm_setzero_pd();
164 fiz2 = _mm_setzero_pd();
166 /* Reset potential sums */
167 velecsum = _mm_setzero_pd();
168 vvdwsum = _mm_setzero_pd();
170 /* Start inner kernel loop */
171 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
174 /* Get j neighbor index, and coordinate index */
177 j_coord_offsetA = DIM*jnrA;
178 j_coord_offsetB = DIM*jnrB;
180 /* load j atom coordinates */
181 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
184 /* Calculate displacement vector */
185 dx00 = _mm_sub_pd(ix0,jx0);
186 dy00 = _mm_sub_pd(iy0,jy0);
187 dz00 = _mm_sub_pd(iz0,jz0);
188 dx10 = _mm_sub_pd(ix1,jx0);
189 dy10 = _mm_sub_pd(iy1,jy0);
190 dz10 = _mm_sub_pd(iz1,jz0);
191 dx20 = _mm_sub_pd(ix2,jx0);
192 dy20 = _mm_sub_pd(iy2,jy0);
193 dz20 = _mm_sub_pd(iz2,jz0);
195 /* Calculate squared distance and things based on it */
196 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
197 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
198 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
200 rinv00 = gmx_mm_invsqrt_pd(rsq00);
201 rinv10 = gmx_mm_invsqrt_pd(rsq10);
202 rinv20 = gmx_mm_invsqrt_pd(rsq20);
204 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
205 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
206 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
208 /* Load parameters for j particles */
209 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
210 vdwjidx0A = 2*vdwtype[jnrA+0];
211 vdwjidx0B = 2*vdwtype[jnrB+0];
213 fjx0 = _mm_setzero_pd();
214 fjy0 = _mm_setzero_pd();
215 fjz0 = _mm_setzero_pd();
217 /**************************
218 * CALCULATE INTERACTIONS *
219 **************************/
221 /* Compute parameters for interactions between i and j atoms */
222 qq00 = _mm_mul_pd(iq0,jq0);
223 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
224 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
226 /* REACTION-FIELD ELECTROSTATICS */
227 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
228 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
230 /* LENNARD-JONES DISPERSION/REPULSION */
232 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
233 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
234 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
235 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
236 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
238 /* Update potential sum for this i atom from the interaction with this j atom. */
239 velecsum = _mm_add_pd(velecsum,velec);
240 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
242 fscal = _mm_add_pd(felec,fvdw);
244 /* Calculate temporary vectorial force */
245 tx = _mm_mul_pd(fscal,dx00);
246 ty = _mm_mul_pd(fscal,dy00);
247 tz = _mm_mul_pd(fscal,dz00);
249 /* Update vectorial force */
250 fix0 = _mm_add_pd(fix0,tx);
251 fiy0 = _mm_add_pd(fiy0,ty);
252 fiz0 = _mm_add_pd(fiz0,tz);
254 fjx0 = _mm_add_pd(fjx0,tx);
255 fjy0 = _mm_add_pd(fjy0,ty);
256 fjz0 = _mm_add_pd(fjz0,tz);
258 /**************************
259 * CALCULATE INTERACTIONS *
260 **************************/
262 /* Compute parameters for interactions between i and j atoms */
263 qq10 = _mm_mul_pd(iq1,jq0);
265 /* REACTION-FIELD ELECTROSTATICS */
266 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
267 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
269 /* Update potential sum for this i atom from the interaction with this j atom. */
270 velecsum = _mm_add_pd(velecsum,velec);
274 /* Calculate temporary vectorial force */
275 tx = _mm_mul_pd(fscal,dx10);
276 ty = _mm_mul_pd(fscal,dy10);
277 tz = _mm_mul_pd(fscal,dz10);
279 /* Update vectorial force */
280 fix1 = _mm_add_pd(fix1,tx);
281 fiy1 = _mm_add_pd(fiy1,ty);
282 fiz1 = _mm_add_pd(fiz1,tz);
284 fjx0 = _mm_add_pd(fjx0,tx);
285 fjy0 = _mm_add_pd(fjy0,ty);
286 fjz0 = _mm_add_pd(fjz0,tz);
288 /**************************
289 * CALCULATE INTERACTIONS *
290 **************************/
292 /* Compute parameters for interactions between i and j atoms */
293 qq20 = _mm_mul_pd(iq2,jq0);
295 /* REACTION-FIELD ELECTROSTATICS */
296 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
297 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
299 /* Update potential sum for this i atom from the interaction with this j atom. */
300 velecsum = _mm_add_pd(velecsum,velec);
304 /* Calculate temporary vectorial force */
305 tx = _mm_mul_pd(fscal,dx20);
306 ty = _mm_mul_pd(fscal,dy20);
307 tz = _mm_mul_pd(fscal,dz20);
309 /* Update vectorial force */
310 fix2 = _mm_add_pd(fix2,tx);
311 fiy2 = _mm_add_pd(fiy2,ty);
312 fiz2 = _mm_add_pd(fiz2,tz);
314 fjx0 = _mm_add_pd(fjx0,tx);
315 fjy0 = _mm_add_pd(fjy0,ty);
316 fjz0 = _mm_add_pd(fjz0,tz);
318 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
320 /* Inner loop uses 111 flops */
327 j_coord_offsetA = DIM*jnrA;
329 /* load j atom coordinates */
330 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
333 /* Calculate displacement vector */
334 dx00 = _mm_sub_pd(ix0,jx0);
335 dy00 = _mm_sub_pd(iy0,jy0);
336 dz00 = _mm_sub_pd(iz0,jz0);
337 dx10 = _mm_sub_pd(ix1,jx0);
338 dy10 = _mm_sub_pd(iy1,jy0);
339 dz10 = _mm_sub_pd(iz1,jz0);
340 dx20 = _mm_sub_pd(ix2,jx0);
341 dy20 = _mm_sub_pd(iy2,jy0);
342 dz20 = _mm_sub_pd(iz2,jz0);
344 /* Calculate squared distance and things based on it */
345 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
346 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
347 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
349 rinv00 = gmx_mm_invsqrt_pd(rsq00);
350 rinv10 = gmx_mm_invsqrt_pd(rsq10);
351 rinv20 = gmx_mm_invsqrt_pd(rsq20);
353 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
354 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
355 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
357 /* Load parameters for j particles */
358 jq0 = _mm_load_sd(charge+jnrA+0);
359 vdwjidx0A = 2*vdwtype[jnrA+0];
361 fjx0 = _mm_setzero_pd();
362 fjy0 = _mm_setzero_pd();
363 fjz0 = _mm_setzero_pd();
365 /**************************
366 * CALCULATE INTERACTIONS *
367 **************************/
369 /* Compute parameters for interactions between i and j atoms */
370 qq00 = _mm_mul_pd(iq0,jq0);
371 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
373 /* REACTION-FIELD ELECTROSTATICS */
374 velec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_add_pd(rinv00,_mm_mul_pd(krf,rsq00)),crf));
375 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
377 /* LENNARD-JONES DISPERSION/REPULSION */
379 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
380 vvdw6 = _mm_mul_pd(c6_00,rinvsix);
381 vvdw12 = _mm_mul_pd(c12_00,_mm_mul_pd(rinvsix,rinvsix));
382 vvdw = _mm_sub_pd( _mm_mul_pd(vvdw12,one_twelfth) , _mm_mul_pd(vvdw6,one_sixth) );
383 fvdw = _mm_mul_pd(_mm_sub_pd(vvdw12,vvdw6),rinvsq00);
385 /* Update potential sum for this i atom from the interaction with this j atom. */
386 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
387 velecsum = _mm_add_pd(velecsum,velec);
388 vvdw = _mm_unpacklo_pd(vvdw,_mm_setzero_pd());
389 vvdwsum = _mm_add_pd(vvdwsum,vvdw);
391 fscal = _mm_add_pd(felec,fvdw);
393 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
395 /* Calculate temporary vectorial force */
396 tx = _mm_mul_pd(fscal,dx00);
397 ty = _mm_mul_pd(fscal,dy00);
398 tz = _mm_mul_pd(fscal,dz00);
400 /* Update vectorial force */
401 fix0 = _mm_add_pd(fix0,tx);
402 fiy0 = _mm_add_pd(fiy0,ty);
403 fiz0 = _mm_add_pd(fiz0,tz);
405 fjx0 = _mm_add_pd(fjx0,tx);
406 fjy0 = _mm_add_pd(fjy0,ty);
407 fjz0 = _mm_add_pd(fjz0,tz);
409 /**************************
410 * CALCULATE INTERACTIONS *
411 **************************/
413 /* Compute parameters for interactions between i and j atoms */
414 qq10 = _mm_mul_pd(iq1,jq0);
416 /* REACTION-FIELD ELECTROSTATICS */
417 velec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_add_pd(rinv10,_mm_mul_pd(krf,rsq10)),crf));
418 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
420 /* Update potential sum for this i atom from the interaction with this j atom. */
421 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
422 velecsum = _mm_add_pd(velecsum,velec);
426 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
428 /* Calculate temporary vectorial force */
429 tx = _mm_mul_pd(fscal,dx10);
430 ty = _mm_mul_pd(fscal,dy10);
431 tz = _mm_mul_pd(fscal,dz10);
433 /* Update vectorial force */
434 fix1 = _mm_add_pd(fix1,tx);
435 fiy1 = _mm_add_pd(fiy1,ty);
436 fiz1 = _mm_add_pd(fiz1,tz);
438 fjx0 = _mm_add_pd(fjx0,tx);
439 fjy0 = _mm_add_pd(fjy0,ty);
440 fjz0 = _mm_add_pd(fjz0,tz);
442 /**************************
443 * CALCULATE INTERACTIONS *
444 **************************/
446 /* Compute parameters for interactions between i and j atoms */
447 qq20 = _mm_mul_pd(iq2,jq0);
449 /* REACTION-FIELD ELECTROSTATICS */
450 velec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_add_pd(rinv20,_mm_mul_pd(krf,rsq20)),crf));
451 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
453 /* Update potential sum for this i atom from the interaction with this j atom. */
454 velec = _mm_unpacklo_pd(velec,_mm_setzero_pd());
455 velecsum = _mm_add_pd(velecsum,velec);
459 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
461 /* Calculate temporary vectorial force */
462 tx = _mm_mul_pd(fscal,dx20);
463 ty = _mm_mul_pd(fscal,dy20);
464 tz = _mm_mul_pd(fscal,dz20);
466 /* Update vectorial force */
467 fix2 = _mm_add_pd(fix2,tx);
468 fiy2 = _mm_add_pd(fiy2,ty);
469 fiz2 = _mm_add_pd(fiz2,tz);
471 fjx0 = _mm_add_pd(fjx0,tx);
472 fjy0 = _mm_add_pd(fjy0,ty);
473 fjz0 = _mm_add_pd(fjz0,tz);
475 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
477 /* Inner loop uses 111 flops */
480 /* End of innermost loop */
482 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
483 f+i_coord_offset,fshift+i_shift_offset);
486 /* Update potential energies */
487 gmx_mm_update_1pot_pd(velecsum,kernel_data->energygrp_elec+ggid);
488 gmx_mm_update_1pot_pd(vvdwsum,kernel_data->energygrp_vdw+ggid);
490 /* Increment number of inner iterations */
491 inneriter += j_index_end - j_index_start;
493 /* Outer loop uses 20 flops */
496 /* Increment number of outer iterations */
499 /* Update outer/inner flops */
501 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_VF,outeriter*20 + inneriter*111);
504 * Gromacs nonbonded kernel: nb_kernel_ElecRF_VdwLJ_GeomW3P1_F_sse2_double
505 * Electrostatics interaction: ReactionField
506 * VdW interaction: LennardJones
507 * Geometry: Water3-Particle
508 * Calculate force/pot: Force
511 nb_kernel_ElecRF_VdwLJ_GeomW3P1_F_sse2_double
512 (t_nblist * gmx_restrict nlist,
513 rvec * gmx_restrict xx,
514 rvec * gmx_restrict ff,
515 t_forcerec * gmx_restrict fr,
516 t_mdatoms * gmx_restrict mdatoms,
517 nb_kernel_data_t gmx_unused * gmx_restrict kernel_data,
518 t_nrnb * gmx_restrict nrnb)
520 /* Suffixes 0,1,2,3 refer to particle indices for waters in the inner or outer loop, or
521 * just 0 for non-waters.
522 * Suffixes A,B refer to j loop unrolling done with SSE double precision, e.g. for the two different
523 * jnr indices corresponding to data put in the four positions in the SIMD register.
525 int i_shift_offset,i_coord_offset,outeriter,inneriter;
526 int j_index_start,j_index_end,jidx,nri,inr,ggid,iidx;
528 int j_coord_offsetA,j_coord_offsetB;
529 int *iinr,*jindex,*jjnr,*shiftidx,*gid;
531 real *shiftvec,*fshift,*x,*f;
532 __m128d tx,ty,tz,fscal,rcutoff,rcutoff2,jidxall;
534 __m128d ix0,iy0,iz0,fix0,fiy0,fiz0,iq0,isai0;
536 __m128d ix1,iy1,iz1,fix1,fiy1,fiz1,iq1,isai1;
538 __m128d ix2,iy2,iz2,fix2,fiy2,fiz2,iq2,isai2;
539 int vdwjidx0A,vdwjidx0B;
540 __m128d jx0,jy0,jz0,fjx0,fjy0,fjz0,jq0,isaj0;
541 __m128d dx00,dy00,dz00,rsq00,rinv00,rinvsq00,r00,qq00,c6_00,c12_00;
542 __m128d dx10,dy10,dz10,rsq10,rinv10,rinvsq10,r10,qq10,c6_10,c12_10;
543 __m128d dx20,dy20,dz20,rsq20,rinv20,rinvsq20,r20,qq20,c6_20,c12_20;
544 __m128d velec,felec,velecsum,facel,crf,krf,krf2;
547 __m128d rinvsix,rvdw,vvdw,vvdw6,vvdw12,fvdw,fvdw6,fvdw12,vvdwsum,sh_vdw_invrcut6;
550 __m128d one_sixth = _mm_set1_pd(1.0/6.0);
551 __m128d one_twelfth = _mm_set1_pd(1.0/12.0);
552 __m128d dummy_mask,cutoff_mask;
553 __m128d signbit = gmx_mm_castsi128_pd( _mm_set_epi32(0x80000000,0x00000000,0x80000000,0x00000000) );
554 __m128d one = _mm_set1_pd(1.0);
555 __m128d two = _mm_set1_pd(2.0);
561 jindex = nlist->jindex;
563 shiftidx = nlist->shift;
565 shiftvec = fr->shift_vec[0];
566 fshift = fr->fshift[0];
567 facel = _mm_set1_pd(fr->epsfac);
568 charge = mdatoms->chargeA;
569 krf = _mm_set1_pd(fr->ic->k_rf);
570 krf2 = _mm_set1_pd(fr->ic->k_rf*2.0);
571 crf = _mm_set1_pd(fr->ic->c_rf);
572 nvdwtype = fr->ntype;
574 vdwtype = mdatoms->typeA;
576 /* Setup water-specific parameters */
577 inr = nlist->iinr[0];
578 iq0 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+0]));
579 iq1 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+1]));
580 iq2 = _mm_mul_pd(facel,_mm_set1_pd(charge[inr+2]));
581 vdwioffset0 = 2*nvdwtype*vdwtype[inr+0];
583 /* Avoid stupid compiler warnings */
591 /* Start outer loop over neighborlists */
592 for(iidx=0; iidx<nri; iidx++)
594 /* Load shift vector for this list */
595 i_shift_offset = DIM*shiftidx[iidx];
597 /* Load limits for loop over neighbors */
598 j_index_start = jindex[iidx];
599 j_index_end = jindex[iidx+1];
601 /* Get outer coordinate index */
603 i_coord_offset = DIM*inr;
605 /* Load i particle coords and add shift vector */
606 gmx_mm_load_shift_and_3rvec_broadcast_pd(shiftvec+i_shift_offset,x+i_coord_offset,
607 &ix0,&iy0,&iz0,&ix1,&iy1,&iz1,&ix2,&iy2,&iz2);
609 fix0 = _mm_setzero_pd();
610 fiy0 = _mm_setzero_pd();
611 fiz0 = _mm_setzero_pd();
612 fix1 = _mm_setzero_pd();
613 fiy1 = _mm_setzero_pd();
614 fiz1 = _mm_setzero_pd();
615 fix2 = _mm_setzero_pd();
616 fiy2 = _mm_setzero_pd();
617 fiz2 = _mm_setzero_pd();
619 /* Start inner kernel loop */
620 for(jidx=j_index_start; jidx<j_index_end-1; jidx+=2)
623 /* Get j neighbor index, and coordinate index */
626 j_coord_offsetA = DIM*jnrA;
627 j_coord_offsetB = DIM*jnrB;
629 /* load j atom coordinates */
630 gmx_mm_load_1rvec_2ptr_swizzle_pd(x+j_coord_offsetA,x+j_coord_offsetB,
633 /* Calculate displacement vector */
634 dx00 = _mm_sub_pd(ix0,jx0);
635 dy00 = _mm_sub_pd(iy0,jy0);
636 dz00 = _mm_sub_pd(iz0,jz0);
637 dx10 = _mm_sub_pd(ix1,jx0);
638 dy10 = _mm_sub_pd(iy1,jy0);
639 dz10 = _mm_sub_pd(iz1,jz0);
640 dx20 = _mm_sub_pd(ix2,jx0);
641 dy20 = _mm_sub_pd(iy2,jy0);
642 dz20 = _mm_sub_pd(iz2,jz0);
644 /* Calculate squared distance and things based on it */
645 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
646 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
647 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
649 rinv00 = gmx_mm_invsqrt_pd(rsq00);
650 rinv10 = gmx_mm_invsqrt_pd(rsq10);
651 rinv20 = gmx_mm_invsqrt_pd(rsq20);
653 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
654 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
655 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
657 /* Load parameters for j particles */
658 jq0 = gmx_mm_load_2real_swizzle_pd(charge+jnrA+0,charge+jnrB+0);
659 vdwjidx0A = 2*vdwtype[jnrA+0];
660 vdwjidx0B = 2*vdwtype[jnrB+0];
662 fjx0 = _mm_setzero_pd();
663 fjy0 = _mm_setzero_pd();
664 fjz0 = _mm_setzero_pd();
666 /**************************
667 * CALCULATE INTERACTIONS *
668 **************************/
670 /* Compute parameters for interactions between i and j atoms */
671 qq00 = _mm_mul_pd(iq0,jq0);
672 gmx_mm_load_2pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,
673 vdwparam+vdwioffset0+vdwjidx0B,&c6_00,&c12_00);
675 /* REACTION-FIELD ELECTROSTATICS */
676 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
678 /* LENNARD-JONES DISPERSION/REPULSION */
680 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
681 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
683 fscal = _mm_add_pd(felec,fvdw);
685 /* Calculate temporary vectorial force */
686 tx = _mm_mul_pd(fscal,dx00);
687 ty = _mm_mul_pd(fscal,dy00);
688 tz = _mm_mul_pd(fscal,dz00);
690 /* Update vectorial force */
691 fix0 = _mm_add_pd(fix0,tx);
692 fiy0 = _mm_add_pd(fiy0,ty);
693 fiz0 = _mm_add_pd(fiz0,tz);
695 fjx0 = _mm_add_pd(fjx0,tx);
696 fjy0 = _mm_add_pd(fjy0,ty);
697 fjz0 = _mm_add_pd(fjz0,tz);
699 /**************************
700 * CALCULATE INTERACTIONS *
701 **************************/
703 /* Compute parameters for interactions between i and j atoms */
704 qq10 = _mm_mul_pd(iq1,jq0);
706 /* REACTION-FIELD ELECTROSTATICS */
707 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
711 /* Calculate temporary vectorial force */
712 tx = _mm_mul_pd(fscal,dx10);
713 ty = _mm_mul_pd(fscal,dy10);
714 tz = _mm_mul_pd(fscal,dz10);
716 /* Update vectorial force */
717 fix1 = _mm_add_pd(fix1,tx);
718 fiy1 = _mm_add_pd(fiy1,ty);
719 fiz1 = _mm_add_pd(fiz1,tz);
721 fjx0 = _mm_add_pd(fjx0,tx);
722 fjy0 = _mm_add_pd(fjy0,ty);
723 fjz0 = _mm_add_pd(fjz0,tz);
725 /**************************
726 * CALCULATE INTERACTIONS *
727 **************************/
729 /* Compute parameters for interactions between i and j atoms */
730 qq20 = _mm_mul_pd(iq2,jq0);
732 /* REACTION-FIELD ELECTROSTATICS */
733 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
737 /* Calculate temporary vectorial force */
738 tx = _mm_mul_pd(fscal,dx20);
739 ty = _mm_mul_pd(fscal,dy20);
740 tz = _mm_mul_pd(fscal,dz20);
742 /* Update vectorial force */
743 fix2 = _mm_add_pd(fix2,tx);
744 fiy2 = _mm_add_pd(fiy2,ty);
745 fiz2 = _mm_add_pd(fiz2,tz);
747 fjx0 = _mm_add_pd(fjx0,tx);
748 fjy0 = _mm_add_pd(fjy0,ty);
749 fjz0 = _mm_add_pd(fjz0,tz);
751 gmx_mm_decrement_1rvec_2ptr_swizzle_pd(f+j_coord_offsetA,f+j_coord_offsetB,fjx0,fjy0,fjz0);
753 /* Inner loop uses 91 flops */
760 j_coord_offsetA = DIM*jnrA;
762 /* load j atom coordinates */
763 gmx_mm_load_1rvec_1ptr_swizzle_pd(x+j_coord_offsetA,
766 /* Calculate displacement vector */
767 dx00 = _mm_sub_pd(ix0,jx0);
768 dy00 = _mm_sub_pd(iy0,jy0);
769 dz00 = _mm_sub_pd(iz0,jz0);
770 dx10 = _mm_sub_pd(ix1,jx0);
771 dy10 = _mm_sub_pd(iy1,jy0);
772 dz10 = _mm_sub_pd(iz1,jz0);
773 dx20 = _mm_sub_pd(ix2,jx0);
774 dy20 = _mm_sub_pd(iy2,jy0);
775 dz20 = _mm_sub_pd(iz2,jz0);
777 /* Calculate squared distance and things based on it */
778 rsq00 = gmx_mm_calc_rsq_pd(dx00,dy00,dz00);
779 rsq10 = gmx_mm_calc_rsq_pd(dx10,dy10,dz10);
780 rsq20 = gmx_mm_calc_rsq_pd(dx20,dy20,dz20);
782 rinv00 = gmx_mm_invsqrt_pd(rsq00);
783 rinv10 = gmx_mm_invsqrt_pd(rsq10);
784 rinv20 = gmx_mm_invsqrt_pd(rsq20);
786 rinvsq00 = _mm_mul_pd(rinv00,rinv00);
787 rinvsq10 = _mm_mul_pd(rinv10,rinv10);
788 rinvsq20 = _mm_mul_pd(rinv20,rinv20);
790 /* Load parameters for j particles */
791 jq0 = _mm_load_sd(charge+jnrA+0);
792 vdwjidx0A = 2*vdwtype[jnrA+0];
794 fjx0 = _mm_setzero_pd();
795 fjy0 = _mm_setzero_pd();
796 fjz0 = _mm_setzero_pd();
798 /**************************
799 * CALCULATE INTERACTIONS *
800 **************************/
802 /* Compute parameters for interactions between i and j atoms */
803 qq00 = _mm_mul_pd(iq0,jq0);
804 gmx_mm_load_1pair_swizzle_pd(vdwparam+vdwioffset0+vdwjidx0A,&c6_00,&c12_00);
806 /* REACTION-FIELD ELECTROSTATICS */
807 felec = _mm_mul_pd(qq00,_mm_sub_pd(_mm_mul_pd(rinv00,rinvsq00),krf2));
809 /* LENNARD-JONES DISPERSION/REPULSION */
811 rinvsix = _mm_mul_pd(_mm_mul_pd(rinvsq00,rinvsq00),rinvsq00);
812 fvdw = _mm_mul_pd(_mm_sub_pd(_mm_mul_pd(c12_00,rinvsix),c6_00),_mm_mul_pd(rinvsix,rinvsq00));
814 fscal = _mm_add_pd(felec,fvdw);
816 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
818 /* Calculate temporary vectorial force */
819 tx = _mm_mul_pd(fscal,dx00);
820 ty = _mm_mul_pd(fscal,dy00);
821 tz = _mm_mul_pd(fscal,dz00);
823 /* Update vectorial force */
824 fix0 = _mm_add_pd(fix0,tx);
825 fiy0 = _mm_add_pd(fiy0,ty);
826 fiz0 = _mm_add_pd(fiz0,tz);
828 fjx0 = _mm_add_pd(fjx0,tx);
829 fjy0 = _mm_add_pd(fjy0,ty);
830 fjz0 = _mm_add_pd(fjz0,tz);
832 /**************************
833 * CALCULATE INTERACTIONS *
834 **************************/
836 /* Compute parameters for interactions between i and j atoms */
837 qq10 = _mm_mul_pd(iq1,jq0);
839 /* REACTION-FIELD ELECTROSTATICS */
840 felec = _mm_mul_pd(qq10,_mm_sub_pd(_mm_mul_pd(rinv10,rinvsq10),krf2));
844 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
846 /* Calculate temporary vectorial force */
847 tx = _mm_mul_pd(fscal,dx10);
848 ty = _mm_mul_pd(fscal,dy10);
849 tz = _mm_mul_pd(fscal,dz10);
851 /* Update vectorial force */
852 fix1 = _mm_add_pd(fix1,tx);
853 fiy1 = _mm_add_pd(fiy1,ty);
854 fiz1 = _mm_add_pd(fiz1,tz);
856 fjx0 = _mm_add_pd(fjx0,tx);
857 fjy0 = _mm_add_pd(fjy0,ty);
858 fjz0 = _mm_add_pd(fjz0,tz);
860 /**************************
861 * CALCULATE INTERACTIONS *
862 **************************/
864 /* Compute parameters for interactions between i and j atoms */
865 qq20 = _mm_mul_pd(iq2,jq0);
867 /* REACTION-FIELD ELECTROSTATICS */
868 felec = _mm_mul_pd(qq20,_mm_sub_pd(_mm_mul_pd(rinv20,rinvsq20),krf2));
872 fscal = _mm_unpacklo_pd(fscal,_mm_setzero_pd());
874 /* Calculate temporary vectorial force */
875 tx = _mm_mul_pd(fscal,dx20);
876 ty = _mm_mul_pd(fscal,dy20);
877 tz = _mm_mul_pd(fscal,dz20);
879 /* Update vectorial force */
880 fix2 = _mm_add_pd(fix2,tx);
881 fiy2 = _mm_add_pd(fiy2,ty);
882 fiz2 = _mm_add_pd(fiz2,tz);
884 fjx0 = _mm_add_pd(fjx0,tx);
885 fjy0 = _mm_add_pd(fjy0,ty);
886 fjz0 = _mm_add_pd(fjz0,tz);
888 gmx_mm_decrement_1rvec_1ptr_swizzle_pd(f+j_coord_offsetA,fjx0,fjy0,fjz0);
890 /* Inner loop uses 91 flops */
893 /* End of innermost loop */
895 gmx_mm_update_iforce_3atom_swizzle_pd(fix0,fiy0,fiz0,fix1,fiy1,fiz1,fix2,fiy2,fiz2,
896 f+i_coord_offset,fshift+i_shift_offset);
898 /* Increment number of inner iterations */
899 inneriter += j_index_end - j_index_start;
901 /* Outer loop uses 18 flops */
904 /* Increment number of outer iterations */
907 /* Update outer/inner flops */
909 inc_nrnb(nrnb,eNR_NBKERNEL_ELEC_VDW_W3_F,outeriter*18 + inneriter*91);